Task Knowledge Research Articles

Multiagent Reinforcement Learning for Strategic Decision Making and Control in Robotic Soccer Through Self-Play

Reinforcement Learning (RL) has shown promising performance in environments for both robotic control and strategic decision making. However, they are usually treated as separate problems with different objectives. In this work, we propose the use of Reinforcement Learning to solve both control and strategic problems as one, in a multi-agent robotic soccer environment. We use the IEEE Very Small Size Soccer (VSSS) challenge from the Latin American Robotics Competition (LARC) as a study case. In the VSSS, two autonomous teams of wheeled robots compete by pushing the ball around to score goals. To unify both control and strategy problems, our approach gives full control of the actuators’ speed to the RL algorithm whilst keeping the broader objective of winning the game. Our method achieves win rates as high as 93% against hand-coded heuristic strategies. In this work we contribute by developing an RL agent that can learn from self-play and generalize against new opponents. Our methodology uses multi-agent Reinforcement Learning with self-play in order to build up the knowledge for complex tasks. We also developed a simulated environment for the robotic soccer game.

Distributed Location-Aware Task Offloading in Multi-UAVs Enabled Edge Computing

Edge computing is a new computing paradigm which distributes tasks to edge networks for processing, it provides effective support for various mobile applications to meet their rapid response requirement. Unmanned Aerial Vehicle (UAV) has been widely used in emergency rescue, mapping, etc., with the advantages of flexible deployment and rapid movement. However, on one hand, mobile applications will be terminated when the battery energy is exhausted. On the other hand, mobile applications will be out of service when mobile devices are out of radio coverage of UAVs. How to achieve low-cost task unloading in the resource limited and location sensitive multi-UAVs edge computing environment is rather challenging. In this paper, we propose a distributed location-aware task offloading scheme, aiming at the above issues. Specifically, we create a nonlinear task allocation problem by combining the limited energy constraints of edge nodes with the random movement of users, where the cost function is divided into static and dynamic costs, respectively. Then, we formulate this problem to a convex optimization one with linear constrains, based on regularization technology. The mathematical proof shows that the scheme can support a parameterized competitive ratio without requiring any prior knowledge of the input task. The simulation results show that the proposed scheme can achieve lower cost edge computing services.

Categorical-ontological approach to information support of educational activities

An important basis for information support of educational activities is the effective presentation of knowledge and standardization of training tasks based on ontological models of learned subject areas. The competence approach to the organization of the educational process and the requirements for the training of specialists from the point view of enterprises and organizations, in which they will carry out their professional activities, requires the development of a comprehensive model of educational processes and objects that are participating in them, based on high-level abstractions. On the basis of the results obtained in this work, the methodology of informational support of educational activities was further developed on the basis of: organizing the educational process and standardization of education based on a categorical-ontological approach; construction and use of relevant on tologies and knowledge bases; automation of data processing by forming and executing queries to the content of ontological models. The use of the results of categorical-ontological modeling allowed to reconcile the interests of active agents of the organization-technical systems, the subjects of educational activity and the ability of students to acquire the necessary knowledge and skills.

Complementary learning-team machines to enlighten and exploit human expertise

The benefits of Industry 4.0 are limited by the large computational requirements of ever-larger digital models of complex production systems. A complementary learning paradigm is thus proposed to cultivate knowledge in a team of machines and humans that represents the key to a high-performance manufacturing system. Two types of knowledge are created using light-weighted neural networks and meta-learning: general knowledge of tasks and specific knowledge on collaboration with humans given few interactions. AI-based teaming strategies are designed to enable machines to leverage human expertise in making decisions using local communications that make intricate sensor systems and expensive computation unnecessary.

System of Robot Learning from Multi-Modal Demonstration and Natural Language Instruction

Collaborative robots are set to play an important role in the future of the manufacturing industry. They need to be able to work outside of the fencing and perform new tasks to individual customer specifications. The necessity of frequent robot re-programming is a great challenge for small and medium sized companies alike. Learning from demonstration is a promising approach that aims to enable robots to acquire from their end users new task knowledge consisting of a sequence of actions, the associated skills, and the context in which the task is executed. Current systems have limited support for integrating semantics and environmental changes. This paper introduces a system combining several modalities as demonstration interfaces, including natural language instruction, visual observation and hand-guiding, which enables the robot to learn a task comprising a goal concept, a plan and basic actions, with consideration for the current environment state. The task thus learned can then be generalized to similar tasks involving different initial and goal states.

Practical subjects in the vocational curriculum: A critical review of the literature

There is a growing number of studies in the field of vocational education and training, but not many focus on the practical subjects that prepare students for transition to workplaces. This critical review of the literature on practical subjects in vocational colleges set out to contribute to debates on the relationship between theory and practice in vocational education through a particular focus on the knowledge forms that are evident in curricular and pedagogical practices in practical subjects. The critical review identified the conceptual, procedural, technical, and contextual forms of knowledge in practical tasks and activities. Based on these findings and drawing on the concept of semantic gravity in Legitimation Code Theory, we found stronger and weaker forms of semantic gravity in practical knowledge and propose that strengthening and weakening semantic gravity through the purposeful selection and sequencing of tasks and activities enables cumulative knowledge building towards competent practice.

Visual domain knowledge-based multimodal zoning for textual region localization in noisy historical document images

Document layout analysis, or zoning, is important for textual content analysis such as optical character recognition. Zoning document images such as digitized historical newspaper pages are challenging due to noise and quality of the document images. Recently, effective data-driven approaches, such as leveraging deep learning, have been proposed, albeit with the concern of requiring larger training data and thus incurring additional cost of ground truthing. We propose a zoning solution by incorporating a knowledge-driven document representation, gravity map, into a multimodal deep learning framework to reduce the amount of time and data required for training. We first generate a gravity map for each image, considering the centroid distance and area between a cell in a Voronoi tessellation and its content to encode visual domain knowledge of a zoning task. Second, we inject the gravity maps into a deep convolution neural network (DCNN) during training, as an additional modality to boost performance. We report on two investigations using two state-of-the-art DCNN architectures and three datasets: two sets of historical newspapers and a set of born-digital contemporary documents. Evaluations show that our solution achieved comparable segmentation accuracy using fewer training epochs and less training data compared to a naïve training scheme.

Rapidly Learning Generalizable and Robot-Agnostic Tool-Use Skills for a Wide Range of Tasks.

Many real-world applications require robots to use tools. However, robots lack the skills necessary to learn and perform many essential tool-use tasks. To this end, we present the TRansferrIng Skilled Tool Use Acquired Rapidly (TRI-STAR) framework for task-general robot tool use. TRI-STAR has three primary components: 1) the ability to learn and apply tool-use skills to a wide variety of tasks from a minimal number of training demonstrations, 2) the ability to generalize learned skills to other tools and manipulated objects, and 3) the ability to transfer learned skills to other robots. These capabilities are enabled by TRI-STAR’s task-oriented approach, which identifies and leverages structural task knowledge through the use of our goal-based task taxonomy. We demonstrate this framework with seven tasks that impose distinct requirements on the usages of the tools, six of which were each performed on three physical robots with varying kinematic configurations. Our results demonstrate that TRI-STAR can learn effective tool-use skills from only 20 training demonstrations. In addition, our framework generalizes tool-use skills to morphologically distinct objects and transfers them to new platforms, with minor performance degradation.

Phase-Synchronized Stimulus Presentation Augments Contingency Knowledge and Affective Evaluation in a Fear-Conditioning Task.

Memory often combines information from different sensory modalities. Animal studies show that synchronized neuronal activity in the theta band (4–8 Hz) binds multimodal associations. Studies with human participants have likewise established that theta-phase synchronization augments the formation of declarative video–tone pair memories. Another form of associative learning, classical fear conditioning, models nondeclarative, emotional memory with distinct neuronal mechanisms. Typical fear-conditioning tasks pair a conditioned stimulus (CS) in one modality with an aversive unconditioned stimulus (US) in another. The present study examines the effects of CS–US synchronization in the theta band on fear memory formation in humans. In a fear generalization procedure, we paired one of five visual gratings of varying orientation (CS) with an aversive auditory US. We modulated the luminance of the CS and the volume of the US at a rate of 4 Hz. To manipulate the synchrony between visual and auditory input during fear acquisition, one group (N = 20) received synchronous CS–US pairing, whereas the control group (N = 20) received the CS–US pairs out of phase. Phase synchronization improved CS–US contingency knowledge and facilitated CS discrimination in terms of rated valence and arousal, resulting in narrower generalization across the CS gratings compared with the out-of-phase group. In contrast, synchronization did not amplify conditioned responding in physiological arousal (skin conductance) and visuocortical engagement (steady-state visually evoked potentials) during acquisition, although both measures demonstrated tuning toward the CS+. Together, these data support a causal role of theta-phase synchronization in affective evaluation and contingency report during fear acquisition.

Analysis of Fiber-Reinforced Concrete Slabs under Centric and Eccentric Load.

Research on the interaction between slabs and subsoil involves the field of materials engineering, concrete structures, and geotechnics. In the vast majority of cases, research focuses on only one of these areas, whereas for advanced study and computer simulations, detailed knowledge of the whole task is required. Among the new knowledge and information upon which this article focuses is the evaluation of subsoil stress using specialized pressure cells, along with detailed measurements of the deformation of a fiber-reinforced concrete slab. From a design point of view, this research is focused on the issue of the center of the cross section and the influence of eccentricity. Knowledge in this area is not yet comprehensively available for fiber-reinforced concrete slabs, where 2D deformation sections of the slab and 3D deformation surfaces of the slab are used in experiments. The experimental program includes a centrically and eccentrically loaded slab. These are structural elements that were tested on a specialized device. Both slabs had the same concrete recipe, with a dispersed reinforcement content of 25 kg/m3. The dimensions of the slab were 2000 × 2000 × 150 mm. Laboratory tests assessed compressive strength, the modulus of elasticity, splitting tensile strength, and bending tensile strength. Based on approximate data from the 3D deformation surfaces, an evaluation of the load-displacement diagrams for the center of the slab and for the center of eccentricity was performed. In conclusion, an overall evaluation and discussion of the results relies on experiments and the mechanical properties of fiber-reinforced concrete.

An expectation-performance gap between money laundering reporting officers and their directors – evidence from Germany

PurposeMoney Laundering Reporting Officers (MLROs) carry out day-to-day anti-money laundering (AML) tasks while directors ultimately remain responsible for AML compliance. Therefore, directors’ expectations of what their MLROs do should ideally coincide with what their actual tasks to minimise liability risk. This paper aims to test for gaps between MLROs and their directors in terms of knowledge, expectations and performance of AML tasks. Likewise, it is researched whether MLROs and directors communicate well with regard to MLROs’ tasks.Design/methodology/approachThis paper first develops a model for analysing the dyadic relationship between MLROs and their directors, based on the audit expectation-performance gap. Second, a paired electronic survey of MLROs and directors of German companies was conducted in autumn 2020, testing for participants’ knowledge, expectations and performance of possible AML tasks (n = 136 pairs).FindingsWhile there is no knowledge or performance gap among MLROs and directors, expectations among them are partially unreasonable and their communication needs to be improved. Additionally, this study suggests that MLROs of German non-financial businesses are less knowledgeable, perform AML duties more poorly, and communicate less effectively with their directors.Practical implicationsTraining of MLROs and communication with their directors need to be improved. Especially in the non-financial sector, action is urgently required.Originality/valueThis paper reports the results of the first paired survey of MLROs and their directors, offering unique insights into their relationship and the status of private AML efforts.

Evaluating the learning and performance characteristics of self-organizing systems with different task features

AbstractSelf-organizing systems (SOS) are developed to perform complex tasks in unforeseen situations with adaptability. Predefining rules for self-organizing agents can be challenging, especially in tasks with high complexity and changing environments. Our previous work has introduced a multiagent reinforcement learning (RL) model as a design approach to solving the rule generation problem of SOS. A deep multiagent RL algorithm was devised to train agents to acquire the task and self-organizing knowledge. However, the simulation was based on one specific task environment. Sensitivity of SOS to reward functions and systematic evaluation of SOS designed with multiagent RL remain an issue. In this paper, we introduced a rotation reward function to regulate agent behaviors during training and tested different weights of such reward on SOS performance in two case studies: box-pushing and T-shape assembly. Additionally, we proposed three metrics to evaluate the SOS: learning stability, quality of learned knowledge, and scalability. Results show that depending on the type of tasks; designers may choose appropriate weights of rotation reward to obtain the full potential of agents’ learning capability. Good learning stability and quality of knowledge can be achieved with an optimal range of team sizes. Scaling up to larger team sizes has better performance than scaling downwards.

Ability to Consolidate Instances as a Proxy for the Association Among Reading, Spelling, and Math Learning Skill.

Learning skills (as well as disorders) tend to be associated; however, cognitive models typically focus either on reading, spelling or maths providing no clear basis for interpreting this phenomenon. A recent new model of learning cognitive skills proposes that the association among learning skills (and potentially the comorbidity of learning disorders) depends in part from the individual ability to consolidate instances (taken as a measure of rate of learning). We examined the performance of typically developing fifth graders over the acquisition of a novel paper-and-pencil task that could be solved based on an algorithm or, with practice, with reference to specific instances. Our aim was to establish a measure of individual rate of learning using parameters envisaged by the instance theory of automatization by Logan and correlate it to tasks requiring knowledge of individual items (e.g., spelling words with an ambiguous transcription) or tasks requiring the application of a rule or an algorithm (e.g., spelling non-words). The paper-and-pencil procedure yielded acquisition curves consistent with the predictions of the instance theory of automatization (i.e., they followed a power function fit) both at a group and an individual level. Performance in tasks requiring knowledge of individual items (such as doing tables or the retrieval of lexical representations) but not in tasks requiring the application of rules or algorithms (such as judging numerosity or spelling through sublexical mapping) was significantly predicted by the learning parameters of the individual power fits. The results support the hypothesis that an individual dimension of “ability to consolidate instances” contributes to learning skills such as reading, spelling, and maths, providing an interesting heuristic for understanding the comorbidity across learning disorders.

Prior task experience increases 5-year-old children's use of proactive control: Behavioral and pupillometric evidence.

Children engage cognitive control reactively when they encounter conflicts; however, they can also resolve conflicts proactively. Recent studies have begun to clarify the mechanisms that support the use of proactive control in children; nonetheless, sufficient knowledge has not been accumulated regarding these mechanisms. Using behavioral and pupillometric measures, we tested the novel possibility that 5-year-old children (N=58) learn to use proactive control via the acquisition of abstract task knowledge that captures regularities of the task. Participants were assigned to either a proactive training group or a control training group. In the proactive training group, participants engaged in a training phase where using proactive control was encouraged, followed by a test phase using different stimuli where both proactive and reactive control could be used. In the control training group, participants engaged in a training phase where both cognitive control strategies could be used, followed by a similarly-structured test phase using different stimuli. We demonstrated children in the control training group responded more quickly and accurately and showed greater cue-related pupil dilation in the test phase than in the training phase. However, there were no differences in response times, accuracies, and pupil dilation between the proactive and control training groups in the training and test phases. These findings suggest that prior task experience, that goes beyond specific knowledge about the timing of task goal activation, can lead children to engage more proactive control endogenously, even if they are not directly encouraged to do so.

A digital twin-driven human–robot collaborative assembly-commissioning method for complex products

The process of complex product assembly-commissioning has the characteristics of high flexibility and firm dynamics. To overcome the drawbacks of manual assembly, deploying automated and intelligent techniques can greatly boost efficiency, improve flexibility, and enhance the quality control. The human–robot collaborative (HRC) technology combines the advantages of human capabilities and the efficiency and precision of robots. However, current HRC technology lacks of perception and cognitive ability, especially in dynamic environments. Therefore, this paper proposed a digital twin-driven HRC assembly-commissioning framework. In this framework, a virtual-physical mapping environment for HRC is constructed. In order to improve the cognitive ability of robot units to tasks, a motion intention recognition approach is proposed which integrates the feature of part into human joint sequences. To improve the adaptability of the robot unit to tasks, the assembly-commissioning task knowledge graph is developed to extract the action sequence of the robot unit in a timely manner. Moreover, the deep deterministic policy gradient (DDPG) is used to adaptively adjust the robot unit movement path in the process of assembly-commissioning. Finally, the effectiveness of the proposed method is verified by taking a particular type of automobile generator as a case study product.

Exploring the Experiences of Student Volunteer and Student Volunteer Chair Communities at Academic Conferences

We report on an interview study with two groups, student volunteers (SV) and student volunteer chairs (SVC), to explore their experiences in contributing to an academic conference. We identified themes regarding their behaviors and perceptions when communicating with each other through an open coding approach. The analysis of the findings reveals that communication issues exist between SV and SVC. While SV faced challenges approaching SVC, SVC had experienced difficulties in finding the right SV to ask for assistance in the tasks. We also found that the information provided by SVC was not sufficient for SV to perform the assigned tasks. Similarly, SVC had a similar issue where they received limited information from former SVC. We present design implications that enable both SV and SVC to share their availability, preferences, and prior knowledge of potential tasks with each other. This is the first study that reports design opportunities to support SV and SVC.

Argumentation and processing knowledge in open-ended assignment tasks: Challenges and accomplishments among pharmacy students

Students in higher education have been shown to have difficulties in developing their critical thinking skills, such as analysis and problem solving, reasoning and argumentation. Open-ended tasks offer opportunities for students to develop their own interpretations of various sources, to critically analyse domain-specific knowledge and utilize that knowledge in their argumentation. This study focuses on the ability of new Master’s students (n=37) to utilize pharmaceutical knowledge from different sources in producing written arguments and counter-arguments in the context of open-ended assignment task. The data were analysed by qualitative content analysis. The results showed that there was substantial variation in how students analysed and processed pharmaceutical knowledge as well as how they utilized that knowledge in their argumentation. While some students were able to provide comprehensive analysis of the different sources, others superficially analysed and processed the sources and struggled to generate convincing arguments. Students’ written responses were typically one-sided: only a few students provided counter-arguments associated with the pharmaceutical problem-solving situation presented in the task. Understanding the nature of the challenges in argumentation and knowledge processing encountered by pharmacy students can help pharmacy educators to modify their pedagogical practices to better support students’ learning.

A General Method for Transferring Explicit Knowledge into Language Model Pretraining

Recently, pretrained language models, such as Bert and XLNet, have rapidly advanced the state of the art on many NLP tasks. They can model implicit semantic information between words in the text. However, it is solely at the token level without considering the background knowledge. Intuitively, background knowledge influences the efficacy of text understanding. Inspired by this, we focus on improving model pretraining by leveraging external knowledge. Different from recent research that optimizes pretraining models by knowledge masking strategies, we propose a simple but general method to transfer explicit knowledge with pretraining. To be specific, we first match knowledge facts from a knowledge base (KB) and then add a knowledge injunction layer to a transformer directly without changing its architecture. This study seeks to find the direct impact of explicit knowledge on model pretraining. We conduct experiments on 7 datasets using 5 knowledge bases in different downstream tasks. Our investigation reveals promising results in all the tasks. The experiment also verifies that domain-specific knowledge is superior to open-domain knowledge in domain-specific task, and different knowledge bases have different performances in different tasks.

Pairwise Preferences-Based Optimization of a Path-Based Velocity Planner in Robotic Sealing Tasks

Production plants are being re-designed to implement human-centered solutions. Especially considering high added-value operations, robots are required to optimize their behavior to achieve a task quality at least comparable to the one obtained by the skilled operators. A manual programming and tuning of the manipulator is not an efficient solution, requiring to adopt towards automated strategies. Adding external sensors (e.g., cameras) increases the robotic cell complexity and it doesn't solve the issue since it is usually difficult to build explicit reward functions measuring the robot performance, while it is easier for the user to define a qualitative comparison between two experiments. According to these needs, in this letter, the recently-developed preferences-based optimization approach GLISp is employed and adapted to tune the novel developed path-based velocity planner. The implemented solution defines an intuitive human-centered procedure, capable of transferring (through pairwise preferences between experiments) the task knowledge from the operator to the manipulator. A Franka EMIKA panda robot has been employed as a test platform to perform a robotic sealing task (i.e., material deposition task), validating the proposed methodology. The proposed approach has been compared with a programming by demonstration approach, and with the manual tuning of the path-based velocity planner. Achieved results demonstrate the improved deposition quality obtained with the proposed optimized path-based velocity planner methodology in a limited number of experimental trials (20).

Knowledge Management Approaches in Pakistani Universities: Faculty Perceptions

Knowledge management is defined as the process of developing, validating, presenting, sharing, and implementing knowledge. Knowledge management (KM) is a strategy for aiding organisations in identifying, choosing, organising, and disseminating the knowledge and skills necessary for tasks including strategic planning, dynamic learning, problem solving, and decision making. Educational institutions played a significant role in the development, management, and dissemination of information and knowledge among higher education institutions. The study's objectives were to learn about (1) faculty perspectives of existing knowledge management approaches, (2) the advantages of knowledge management practises, and (3) the current hurdles to adopting knowledge management practises effectively inPakistani universities. Participants in the study were all PhD supervisors who had been approved by all public and private institutions in Pakistan. During the interview process, researcher spoke with a total of 15 PhD supervisors. As a basis for qualitative data analysis, grounded theory was employed. Research revealed that Pakistani institutions have a poor understanding of knowledge management. However, despite the fact that universities in Pakistan were well aware of the benefits of knowledge management methods, they did not make any attempts to adopt them. Knowledge management approaches in Pakistan are afflicted by difficulties related to perspectives and ICTs due to a dearth of frameworks, coordination points, and budgetary concerns in the country's institutions. The findings indicated that knowledge management approaches are required in Pakistani universities. It was concluded in this research that knowledge management practices have not been completely included into the planned agendas of most institutions Knowledge management strategies, plans, and policy guidelines are recommended by the study for universities to create